glyph-brush-0.4

• Use queue call counting & fine grained hashing to match up previous calls with current calls figuring out what has changed allowing optimised use of recalculate_glyphs for fast layouts.

  • Compute if just geometry (ie section position) has changed -> GlyphChange::Geometry.
  • Compute if just color has changed -> GlyphChange::Color.
  • Compute if just alpha has changed -> GlyphChange::Alpha.

  Provides much faster re-layout performance in these common change scenarios.

• GlyphBrush now generates & caches vertices avoiding regeneration of individual unchanged sections, when another section change forces regeneration of the complete vertex array. The user vertex type V is now in the struct signature.

  pub struct DownstreamGlyphBrush<'font, H = DefaultSectionHasher> {
      // previously: glyph_brush::GlyphBrush<'font, H>,
      inner: glyph_brush::GlyphBrush<'font, DownstreamGlyphVertex, H>,
      ...
  }

These changes result in a big performance improvement for changes to sections amongst other unchanging sections, which is a fairly common thing to want to do. Fully cached (everything unchanging) & worst-case (everything changing/new) are not significantly affected.

name                                      control ns/iter  change ns/iter  diff ns/iter   diff %  speedup 
continually_modify_alpha_of_1_of_3        854,554          349,165             -505,389  -59.14%   x 2.45 
continually_modify_bounds_of_1_of_3       865,162          643,341             -221,821  -25.64%   x 1.34 
continually_modify_color_of_1_of_3        855,189          344,501             -510,688  -59.72%   x 2.48 
continually_modify_end_text_of_1_of_3     857,913          647,491             -210,422  -24.53%   x 1.32 
continually_modify_middle_text_of_1_of_3  856,383          642,055             -214,328  -25.03%   x 1.33 
continually_modify_position_of_1_of_3     866,298          355,794             -510,504  -58.93%   x 2.43 
continually_modify_start_text_of_1_of_3   872,081          654,887             -217,194  -24.91%   x 1.33
no_cache_render_3_medium_sections_fully   1,480,556        1,500,286             19,730    1.33%   x 0.99
render_3_medium_sections_fully            1,136            1,066                    -70   -6.16%   x 1.07